Nonparametric Identification of Incomplete Information Discrete Games With Non‐Equilibrium Behaviors

This paper studies empirical discrete‐choice games with incomplete information under nonparametric specifications of both the payoff function and the distribution of private information. It relaxes the standard global equilibrium assumption—under which players follow equilibrium strategies for all realizations of the control variables—and introduces a weaker local equilibrium assumption that requires the equilibrium restriction only for some, but not all, realizations. Under this maintained assumption and standard exclusion restrictions, I derive the testable implications of the global equilibrium hypothesis and establish the nonparametric identification of all unknown functions in the model. This paper also discusses settings where the local equilibrium condition applies. The method is illustrated through a Monte Carlo experiment and an empirical application examining the competition between KFC and McDonald's in China. The estimation results strongly reject the global equilibrium assumption.